[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: treatreg

From   Partha Deb <>
Subject   Re: st: treatreg
Date   Sun, 13 Apr 2008 23:17:29 -0400


Two-step fits a probit in the first step so should be identical to the results from a probit. The default (ML) fits a probit-like equation as part of a model with bivariate normal errors. So the treatment equation is not a (univariate normal-based) probit equation. Thus neither coefficients nor predictions will be the same.



Martin Weiss wrote:

Dear Statalisters,

try this in Stata:

webuse labor, clear
generate wc = 0
replace wc = 1 if we > 12

* ML
	treatreg ww wa cit, treat(wc=wmed wfed) first
	predict treatprob, ptrt
	prob wc wmed wfed
	predict probit

compare tre pro
* 2-step
	treatreg ww wa cit, treat(wc=wmed wfed) first two
	predict treatprob2, ptrt

compare treatprob2 pro
l tre* pro in 20/30

and let me know: why are the first stage probit predicitions different
between the ML and the conventional probit but not between the twostep and
the conventional probit? Am I missing something here?

Thanks in advance,
Martin Weiss

Diplom-Kaufmann Martin Weiss
Mohlstrasse 36
Room 415
72074 Tuebingen

Fon: 0049-7071-2978184




*   For searches and help try:
Partha Deb
Department of Economics
Hunter College
ph:  (212) 772-5435
fax: (212) 772-5398

Emancipate yourselves from mental slavery
None but ourselves can free our minds.
	- Bob Marley

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index